hukewei/lerobot_peg_optical_3cams_0525_1

该数据集使用LeRobot创建，包含机器人技术任务的数据。数据集结构基于meta/info.json，具体如下：机器人类型为so_follower，总共有71个episodes，60,868帧，涵盖1个任务。数据以parquet文件存储，总数据文件大小为100MB，视频文件大小为200MB，帧率为30fps。特征包括动作（6个浮点值，对应肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置）、观察状态（与动作相同的6个浮点值）、以及来自三个摄像头的图像观察（左腕、近景和顶部摄像头，每个图像分辨率为480x640，3通道彩色视频，使用av1编解码器）。此外，还包含时间戳、帧索引、episode索引、索引和任务索引等元数据。所有数据按训练分割（0:71）组织。This dataset was created using LeRobot and contains data for robotics tasks. The dataset structure, based on meta/info.json, is as follows: robot type is so_follower, with 71 total episodes, 60,868 frames, and 1 total task. Data is stored in parquet files, with total data file size of 100MB and video file size of 200MB, at 30fps. Features include action (6 float32 values for shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions), observation state (same 6 float32 values), and image observations from three cameras (left wrist, close, and top cameras, each with 480x640 resolution, 3-channel color video, using av1 codec). Additionally, metadata such as timestamp, frame index, episode index, index, and task index are included. All data is organized into a train split (0:71).